https://www.anthropic.com/research/mapping-mind-language-model
Today we report a significant advance in understanding the inner workings of AI models. We have identified how millions of concepts are represented inside Claude Sonnet, one of our deployed large language models. This is the first ever detailed look inside a modern, production-grade large language model. This interpretability discovery could, in future, help us make AI models safer.
Previously, we made some progress matching patterns of neuron activations, called features, to human-interpretable concepts. We used a technique called "dictionary learning", borrowed from classical machine learning, which isolates patterns of neuron activations that recur across many different contexts. In turn, any internal state of the model can be represented in terms of a few active features instead of many active neurons. Just as every English word in a dictionary is made by combining letters, and every sentence is made by combining words, every feature in an AI model is made by combining neurons, and every internal state is made by combining features.
In October 2023, we reported success applying dictionary learning to a very small "toy" language model and found coherent features corresponding to concepts like uppercase text, DNA sequences, surnames in citations, nouns in mathematics, or function arguments in Python code.
We used the same scaling law philosophy that predicts the performance of larger models from smaller ones to tune our methods at an affordable scale before launching on Sonnet.
We successfully extracted millions of features from the middle layer of Claude 3.0 Sonnet, (a member of our current, state-of-the-art model family, currently available on claude.ai), providing a rough conceptual map of its internal states halfway through its computation. This is the first ever detailed look inside a modern, production-grade large language model.